CN113726736B - Identity authentication method and system based on individual behavior cohesion - Google Patents
Identity authentication method and system based on individual behavior cohesion Download PDFInfo
- Publication number
- CN113726736B CN113726736B CN202110838416.0A CN202110838416A CN113726736B CN 113726736 B CN113726736 B CN 113726736B CN 202110838416 A CN202110838416 A CN 202110838416A CN 113726736 B CN113726736 B CN 113726736B
- Authority
- CN
- China
- Prior art keywords
- behavior
- individual
- identity authentication
- cohesiveness
- cohesion
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L63/00—Network architectures or network communication protocols for network security
- H04L63/08—Network architectures or network communication protocols for network security for authentication of entities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Theoretical Computer Science (AREA)
- Computational Mathematics (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Software Systems (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Algebra (AREA)
- Evolutionary Biology (AREA)
- Databases & Information Systems (AREA)
- Probability & Statistics with Applications (AREA)
- Life Sciences & Earth Sciences (AREA)
- Operations Research (AREA)
- Computer Hardware Design (AREA)
- Computer Security & Cryptography (AREA)
- Computing Systems (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Collating Specific Patterns (AREA)
Abstract
The invention relates to an identity authentication method and system based on individual behavior cohesiveness. The method solves the problem of poor individual behavior modeling effect in the prior art. The method comprises the steps of S1, constructing a data table; s2, acquiring individual behavior information from the integrated and intelligent public data platform, and acquiring and storing data; s3, preprocessing the acquired individual behavior information, and S4, building an identity authentication algorithm center based on the cohesiveness of the individual behavior; s5, fully testing, verifying and adjusting the identity authentication algorithm center, and building application on the identity authentication algorithm center; and S6, inputting the behavior information, calculating the behavior information and the stored individual behavior information by the identity authentication algorithm center, outputting the legality probability of the identity authentication, and comparing the legality probability of the output identity authentication with a set threshold value. The invention has the advantages that: the whole individual behaviors are comprehensively depicted, and the modeling effect of the individual behaviors is improved.
Description
Technical Field
The invention relates to the technical field of identity authentication, in particular to an identity authentication method and system based on individual behavior cohesion.
Background
Identity authentication is one of important means for ensuring the security of a network environment, and the existing user identity authentication method, no matter in a single account password mode or a digital certificate mode, has the defects of easy copying, easy stealing, easy propagation and the like which are difficult to overcome by the existing user identity authentication method, and cannot meet the requirements of the security and the uniqueness of identity authentication. The method has the advantages that the inherent physiological characteristics of the human body such as voice, iris, fingerprint and the like are taken as the basis of identity identification, the accuracy rate is objectively high, but additional auxiliary equipment is needed, so the method can only play a role in certain fixed occasions and is not universal. Therefore, in practical application, the effectiveness and reliability of the method are difficult to guarantee.
In contrast, the identity authentication method based on the behaviors has the advantages of non-invasiveness, non-repudiation, stability and the like, is increasingly becoming an important means of identity authentication, depends on mining and depicting individual behavior patterns, effectively reflects the commonalities among individuals while depicting personalized behavior patterns. In practical applications, individual behaviors are characterized collectively by a number of different attributes.
The existing individual behavior modeling method usually considers the behavior attributes separately, and obtains statistical distribution depending on specific attributes by adopting methods such as individual behavior aggregation or feature engineering, and the statistical distribution is used as the measurement of an individual behavior mode. In such a modeling process, the integrity between individual behavior attributes cannot be effectively guaranteed, thereby resulting in poor modeling effect.
Disclosure of Invention
The invention aims to solve the problems and provides an identity authentication method based on individual behavior cohesion, which is reasonable in design and good in using effect.
The invention aims to solve the problems and provides an identity authentication system based on individual behavior cohesion, which is reasonable in design and convenient to operate.
In order to achieve the purpose, the invention adopts the following technical scheme: the identity authentication method based on the individual behavior cohesiveness comprises the following steps:
s1, constructing a data table for recording attribute fields and attribute value types by taking the identity card number of an individual as a main key or an external key; and constructing a data table in advance for preprocessing the acquired individual behavior information.
S2, acquiring individual behavior information from an integrated and intelligent public data platform, and acquiring and storing data by adopting a uniform API (application program interface) or database docking mode; the individual behavior information is collected and stored, and data comparison is convenient to perform in the later stage.
S3, preprocessing the acquired individual behavior information, and forming a data table for completely describing the individual behavior information through preprocessing;
s4, establishing an identity authentication algorithm center based on the individual behavior cohesiveness; the establishment of the identity authentication algorithm center can realize the modeling of individual behavior information, and specifically comprises the establishment of a target function and the design of a training algorithm.
S5, fully testing, verifying and adjusting the identity authentication algorithm center, building application on the identity authentication algorithm center, exposing an interface to the outside, and integrating the interface into an operation link of an integrated and intelligent public data platform needing identity authentication; the identity authentication algorithm center is fully tested, verified and adjusted, so that the generation of algorithm errors can be prevented, and the accuracy of the algorithm is improved.
And S6, inputting the behavior information, carrying out cohesion calculation on the input behavior information by combining the identity authentication algorithm center with the individual behavior model, outputting the validity probability of the identity authentication, comparing the output validity probability of the identity authentication with a set threshold value to obtain an identity authentication result, if the output validity probability of the identity authentication is larger than the set threshold value, indicating that the identity authentication passes, and if the output validity probability of the identity authentication is smaller than the set threshold value, indicating that the identity authentication does not pass. The authentication efficiency is improved through offline model training and online rapid threshold judgment.
In the identity authentication method based on the individual behavior cohesiveness, in step S2, the individual behavior information includes login behavior, browsing behavior, handling behavior and social behavior, wherein,
individual behavior information is composed of a series of attribute characterizations, i.e., τ =<a 1 ，a 2 ，...，a J >Wherein a is j (J =1, 2.. Said., J) is attribute a j Corresponding attribute value, the set of all behaviors is marked as T, and all behavior attribute values are mapped into sharing vector by adopting embedding methodA point in space.
In the identity authentication method based on individual behavior cohesion, in step S3, the preprocessed content includes behavior data table connection, data integrity verification, attribute field normalization processing, continuous field discretization processing, field missing value filling, and invalid data filtering. The individual behavior information is subjected to diversified processing, invalid information is removed, valid information is reserved, data verification and classification are carried out, and the accuracy of the data is improved.
In the identity authentication method based on the individual behavior cohesiveness, in step S4, an identity authentication algorithm center based on the individual behavior cohesiveness is built to include three algorithm module construction and two database construction; the three algorithm modules are respectively an individual behavior cohesiveness modeling algorithm module, a behavior cohesiveness probabilistic algorithm module and an authentication result output module based on threshold judgment; the two databases are an individual behavior library and an attribute embedding library respectively.
In the identity authentication method based on the individual behavior cohesiveness, the individual behavior cohesiveness modeling algorithm module comprises the steps of setting the dimension of an embedded vector space, defining a behavior cohesiveness vector, the individual behavior cohesiveness and cohesiveness partial order, constructing an optimized objective function of the individual behavior cohesiveness, designing a parameter optimization method based on random gradient reduction and carrying out model training, storing the individual behavior vector obtained by training into an individual behavior library, storing the attribute embedded vector obtained by training into an attribute embedded library, wherein,
for behavior Attribute A j (J =1,2, \8230;, J) of any one of the attribute values a j It is mapped to a point in d-dimensional vector space, i.e. to a point in d-dimensional vector space
One behavior τ can be described as an embedded matrix m (τ) = [ a [ ] 1 ，a 2 ，...，a J ]Using the similarity between the embedded matrix column vectors, a cohesive vector for behavior τ can be defined as h (τ) = (a) 1 ·a 2 ，a 1 ·a 3 ，...，a 1 ·a J ，a 2 ·a 3 ，...，a 2 ·a J ，...，a J-1 ·a J ) If the number of attributes describing a behavior is J, then its cohesiveness vector dimension is K = J. (J-1)/2, and accordingly, a K-dimension vector @isused>
Describing the behavior pattern of an individual i, namely a behavior vector of the individual, and defining the cohesiveness of the individual to the behavior on the basis of the behavior vector as follows:
the cohesiveness of an individual i to a behavior tau is defined as its behavior vector b i Inner product with cohesion vector h (τ), i.e.
For an individual i and a behavior τ p ，τ q E.g. T, partial order of cohesiveness
Representing i pairs of behavior τ of individuals p Has a cohesion higher than its behavior τ q Cohesion of (4), if->
Then the behavior τ is passed p Authenticating an individual i to be more legitimacy than the behavior τ q Then the individual's cohesive preference for behavior satisfies the condition: />
If the individual's cohesive preference for behavior is treated as a random event, then the corresponding set of cohesive preference events is: />
Assuming that the cohesive preference events of individual i are independent of each other, the joint probability of all preference events is expressed as: />
Based on preference event->
The probability of its occurrence can be calculated using the following mathematical expression:
wherein it is present>
The function of the method is to convert any real number into a value in an interval (0-1), and further assume that cohesive preference events corresponding to different individuals are independent from each other, an optimized objective function is expressed as: />
Wherein I is the set of all individuals, and theta = { b = i ，v 1 ：，v 2 ：，...，v J : i ∈ I } represents a set of parameters for the model, then the optimal parameters can be obtained by maximizing the above objective function: />
In the above-described identity authentication method based on the individual behavior cohesiveness, for the optimization objective function, the objective function is converted into a negative logarithmic form, i.e., the objective function is converted into a negative logarithmic form 
Wherein it is present>
The parameter lambda controls the influence degree of the regular term, and after the objective function is converted, the optimized parameter can be obtained in the following mode: />
Model parameter optimization by using a stochastic gradient descent method for any cohesive preference->
In terms of this, the gradient of the objective function versus the model parameters is represented as:
wherein, delta = - (1-sigma (b) i ·(h(τ p )-h(τ q ) )) is a constant that depends on the current iteration parameters of the model, a) p，s And a q，s Are respectively the behavior τ p And τ q The s (s =1, 2...., J) th column vector of the embedding matrix, according to the gradient formula, the iterative method of parameters in the training process is as follows:
where η is the learning rate used to control the span of the parameter in each step of the iterative process.
In the identity authentication method based on the individual behavior cohesiveness, the main key of the individual behavior library is the identity card number representing the individual; the primary key of the attribute embedded library is a uniform code containing field names and field value information. This arrangement facilitates subsequent query operations.
In the identity authentication method based on the individual behavior cohesion, the behavior cohesion probabilistic algorithm module comprises the steps of obtaining an individual behavior vector by inquiring from an individual behavior library, obtaining a corresponding vector representation by inquiring from an attribute embedding library according to the field name and the field value of the behavior, calculating the cohesion vector of the behavior, obtaining individual behavior data by calculating and performing probabilistic processing on the individual behavior data according to the individual cohesion definition of the behavior, and outputting the legal probability of identity authentication, wherein,
After the model training is finished, for any individual i, obtaining a behavior vector b representing the behavior pattern of the individual i i For arbitrary behavior τ =<a 1 ，a 2 ，...，a I >Obtaining an embedding matrix m (tau) of the individual behavior, further obtaining an aggregation vector h (tau) of the individual behavior, and correspondingly obtaining the cohesion of the individual behavior
As a basis for the validity of the authentication of an individual i with a behavior τ, however, based on @>
The value (b) is an arbitrary real number, has no fixed value range, and therefore cannot be directly used, and needs to be normalized, that is, the cohesion is based on the individual behavior>
This is translated into a probability of validity to authenticate an individual i with a behavior τ: />
The method comprises the following specific steps: />
A. For all behaviors except τ, [ tau ]' [ epsilon ] T, the behavior cohesiveness of the individual i is calculated separately
And their collection is denoted as F i ；
B. Set counter C, traverse F i Behavior cohesion value of (1)
If->
Then the value of C is added with 1;
C. a probability of validity to authenticate an individual i with a behavior tau is calculated,
in the identity authentication method based on the individual behavior cohesion, the authentication result output module based on threshold judgment comprises threshold setting, and the legality probability of the output identity authentication is compared with the threshold to determine whether the authentication passes, wherein, if the set threshold is p, if so, the authentication passes 
The identity authentication is passed; on the contrary, if +>
The authentication is not passed.
According to the identity authentication method based on the individual behavior cohesiveness, an identity authentication system based on the individual behavior cohesiveness is provided. The identity authentication system has good security.
Compared with the prior art, the invention has the advantages that: the identity authentication method based on the individual behavior cohesiveness is reasonable in design, better in safety, capable of effectively guaranteeing integrity among different individual transaction attributes, capable of improving modeling effect and important in theoretical significance and practical value of the identity authentication based on the behavior.
Drawings
Fig. 1 is a flow chart of identity authentication in the present invention.
Detailed Description
The invention is described in further detail below with reference to the drawings and the detailed description.
As shown in fig. 1, the identity authentication method based on the cohesion of individual behaviors includes the following steps:
s1, constructing a data table for recording attribute fields and attribute value types by taking the identity card number of an individual as a main key or an external key; and sorting and recording the acquired individual behavior data by pre-constructing a data table and determining the main attribute field and attribute value type of the behavior data.
S2, acquiring individual behavior information from an integrated and intelligent public data platform, and acquiring and storing data in a unified API (application program interface) or database docking mode;
s3, preprocessing the acquired individual behavior information, and forming a data table for completely describing the individual behavior information through preprocessing;
s4, establishing an identity authentication algorithm center based on the individual behavior cohesiveness;
s5, fully testing, verifying and adjusting the identity authentication algorithm center, building application on the identity authentication algorithm center, exposing an interface to the outside, and integrating the interface into an operation link of an integrated and intelligent public data platform needing identity authentication; the operation links comprise user login, data application, key resource access and the like, and timely, accurate, non-invasive and personalized identity authentication service can be provided.
And S6, inputting the behavior information, carrying out cohesion calculation on the input behavior information by combining the identity authentication algorithm center with the individual behavior model and outputting the validity probability of identity authentication, comparing the output validity probability of the identity authentication with a set threshold value to obtain an identity authentication result, if the output validity probability of the identity authentication is larger than the set threshold value, the identity authentication is passed, and if the output validity probability of the identity authentication is smaller than the set threshold value, the identity authentication is not passed.
Wherein, in step S2, the individual behavior information includes login behavior, browsing behavior, transaction behavior and social behavior, wherein,
individual behavior information is composed of a series of attribute characterizations, i.e., τ =<a 1 ，a 2 ，...，a J >Wherein a is j = J (J =1, 2.. Gtang., J) is the attribute a j And (4) mapping all behavior attribute values into points in a shared vector space by adopting an embedding method, wherein the set of all behaviors is marked as T. Through data preprocessing, a csv format file which completely describes individual behaviors is formed, and the file can meet the quality requirements of subsequent data analysis links such as machine learning or data mining.
Visibly, in step S3, the preprocessed content includes behavior data table connection, data integrity verification, attribute field normalization processing, continuous field discretization processing, field missing value filling and invalid data filtering.
Obviously, in step S4, the establishment of the identity authentication algorithm center based on the individual behavior cohesiveness includes three algorithm module constructions and two database constructions; the three algorithm modules are respectively an individual behavior cohesiveness modeling algorithm module, a behavior cohesiveness probabilistic algorithm module and an authentication result output module based on threshold judgment; the two databases are an individual behavior library and an attribute embedded library respectively.
Further, the individual behavior cohesiveness modeling algorithm module comprises the steps of setting the dimension of an embedded vector space, defining a behavior cohesiveness vector, individual behavior cohesiveness and cohesiveness partial order, constructing an optimized objective function of the individual behavior cohesiveness, designing a parameter optimization method based on random gradient reduction, performing model training, storing the individual behavior vector obtained through training into an individual behavior library, storing the attribute embedded vector obtained through training into an attribute embedded library, wherein,
for behavior Attribute A j (J =1,2, \8230;, J) of any one of the attribute values a j It is mapped to a point in d-dimensional vector space, i.e. to a point in d-dimensional vector space
One behavior τ can be described as an embedded matrix m (τ) = [ a [ ] 1 ，a 2 ，...，a J ]Using the similarity between embedded matrix column vectors, a cohesive vector for behavior τ can be defined as h (τ) = (a) 1 ·a 2 ，a 1 ·a 3 ，...，a 1 ·a J ，a 2 ·a 3 ，...，a 2 ·a J ，...，a J-1 ·a J ) If the number of attributes describing a behavior is J, then its cohesiveness vector has a dimension of K = J. (J-1)/2, and accordingly, a K-dimensional vector is used>
Describing the behavior pattern of an individual i, namely a behavior vector of the individual, and defining the cohesiveness of the individual to the behavior on the basis of the behavior vector as follows:
the cohesiveness of an individual i to a behavior tau is defined as its behavior vector b i Inner product with cohesive vector h (τ), i.e. 
For an individual i and a behavior τ p ，τ q ∈TPartial order of cohesion
Representing i pairs of behavior τ of individuals p Has a cohesion higher than its behavior τ q Cohesion of (a), if +>
Then the pass behavior τ p Authenticating an individual i as being more legitimate than the behaviour τ q Then the individual's cohesive preference for behavior satisfies the condition: />
If the individual's cohesive preference for behavior is treated as a random event, then the corresponding set of cohesive preference events is: />
Assuming that cohesive preference events for individual i are independent of each other, the joint probability of all preference events is expressed as:
based on preference event->
The probability of its occurrence can be calculated using the following mathematical expression: />
Wherein it is present>
The function of the method is to convert any real number into a value in an interval (0-1), and further assume that cohesive preference events corresponding to different individuals are independent from each other, an optimized objective function is expressed as: />
Wherein I is the set of all individuals, and theta = { b = i ，v 1 ：，v 2 ：，...，v J : i ∈ I } represents a set of parameters for the model, then the optimal parameters can be obtained by maximizing the above objective function: />
The input of the individual behavior cohesiveness modeling algorithm module is collected individual behavior data, and the output is vectorized representation of individual behavior vectors and attribute field values.
In particular, for optimizing the objective function, the objective function is converted into a negative logarithmic form, i.e.
Wherein it is present>
The parameter lambda is a two-norm regular term of the parameter and is used for preventing the over-fitting phenomenon in the training process from occurring, the influence degree of the regular term is controlled by the parameter lambda, and after the objective function is converted, the optimized parameter can be obtained by the following method: />
Optimization of model parameters using stochastic gradient descent for arbitrary cohesive preference>
In terms of this, the gradient of the objective function to the model parameters is represented as:
wherein, δ = - (1- σ (b) i ·(h(τ p )-h(τ q ) )) is a constant that depends on the current iteration parameters of the model, a) p，s And a q，s Are respectively the behavior τ p And τ q The s (s =1, 2...., J) th column vector of the embedding matrix, according to the gradient formula, the iterative method of parameters in the training process is as follows:
where η is the learning rate used to control the span of the parameter in each step of the iterative process. The behavior vectors obtained by training in the individual behavior library can also provide support for other applications or services, such as various personalized recommendation or prediction services.
Furthermore, the main key of the individual behavior library represents the identity card number of the individual; the primary key of the attribute embedded library is a uniform code containing field names and field value information. And the unified coding is set to facilitate subsequent individual behavior information query.
More specifically, the behavior cohesion probabilistic algorithm module comprises a behavior vector of an individual obtained by inquiring from an individual behavior library, a corresponding vector representation obtained by inquiring from an attribute embedding library according to the field name and the field value of the behavior, a cohesion vector of the behavior is calculated, individual behavior data is calculated and subjected to probabilistic processing according to the cohesion definition of the individual to the behavior, and the legal probability of identity authentication is output, wherein,
after the model training is finished, for any individual i, obtaining a behavior vector b representing the behavior pattern of the individual i i For arbitrary behavior τ =<a 1 ，a 2 ，...，a J >Obtaining an embedding matrix m (tau) of the individual behavior, further obtaining an aggregation vector h (tau) of the individual behavior, and correspondingly obtaining the cohesion of the individual behavior
As a basis for the validity of the authentication of an individual i with a behavior τ, however, based on @>
The value (b) is an arbitrary real number, has no fixed value range, and therefore cannot be directly used, and needs to be normalized, that is, the cohesion is based on the individual behavior>
This is translated into a probability of validity to authenticate an individual i with a behavior τ: />
The method comprises the following specific steps:
A. for all behaviors except for τ, [ tau ]' [ epsilon ] T, behavior cohesiveness of the individual i is calculated separately
And their collection is denoted F i ；
B. Set counter C, traverse F i Behavior cohesion value of (1)
If +>
Adding 1 to the C value;
C. a probability of validity to authenticate an individual i with a behavior tau is calculated,
in detail, the authentication result output module based on threshold judgment comprises the setting of a threshold, and whether the authentication passes or not is determined by comparing the validity probability of the output identity authentication with the threshold, wherein the set threshold is assumed to be p, if so, the authentication passes
The identity authentication is passed; on the contrary, if->
The authentication is not passed. The input of the behavior cohesiveness probabilistic algorithm module is an individual i and a behavior τ ', and the output is the validity probability of identity authentication of the individual i according to the behavior τ'.
An identity authentication system based on individual behavior cohesiveness. The objective function is constructed through the cohesion of the individual behaviors and the solving method, so that the overall depiction of the individual behaviors is realized, the modeling effect is improved, and the identity authentication is safer.
In summary, the principle of the present embodiment is: by defining attribute embedding, behavior cohesion vectors, individual cohesion to behaviors, cohesion partial order, construction of a cohesion optimal objective function, a solving method and design of an individual cohesion probabilistic method, the overall description of the individual behaviors is realized, theoretical support and technical realization are provided for solving the sufficiency of individual behavior modeling and the effectiveness of behavior identity authentication, and the modeling effect is improved.
The specific embodiments described herein are merely illustrative of the spirit of the invention. Various modifications or additions may be made to the described embodiments or alternatives may be employed by those skilled in the art without departing from the spirit or ambit of the invention as defined in the appended claims.
Although technical terms such as cohesiveness, individual behavior, etc. are used more herein, the possibility of using other terms is not excluded. These terms are used merely to more conveniently describe and explain the nature of the present invention; they are to be construed as being without limitation to any additional limitations that may be imposed by the spirit of the present invention.
Claims (6)
1. An identity authentication method based on individual behavior cohesion is characterized by comprising the following steps:
s1, constructing a data table for recording attribute fields and attribute value types by taking the identity card number of an individual as a main key or an external key;
s2, acquiring individual behavior information from an integrated and intelligent public data platform, and acquiring and storing data in a unified API (application program interface) or database docking mode;
s3, preprocessing the acquired individual behavior information, and forming a data table for completely describing the individual behavior information through preprocessing;
S4, establishing an identity authentication algorithm center based on the individual behavior cohesiveness;
s5, fully testing, verifying and optimizing the identity authentication algorithm center, building an application on the identity authentication algorithm center, exposing an interface to the outside, and integrating the application into an integrated and intelligent operation link of a public data platform needing identity authentication;
s6, behavior information is input, the identity authentication algorithm center performs cohesion calculation on the input behavior information by combining with the individual behavior model and outputs the validity probability of identity authentication, the output validity probability of the identity authentication is compared with a set threshold value to obtain an identity authentication result, if the output validity probability of the identity authentication is larger than the set threshold value, identity authentication is passed, and if the output validity probability of the identity authentication is smaller than the set threshold value, identity authentication is not passed;
in step S2, the individual behavior information includes login behavior, browsing behavior, transaction behavior and social behavior, wherein,
individual behavior information is composed of a series of attribute characterizations, i.e., τ =<a 1 ，a 2 ，...，a J >Wherein a is j (J =1, 2.. Said., J) is attribute a j Corresponding attribute values, wherein the set of all behaviors is marked as T, and all behavior attribute values are mapped into points in a shared vector space by adopting an embedding method;
In step S3, the preprocessed content includes behavior data table connection, data integrity verification, attribute field normalization processing, continuous field discretization processing, field missing value filling, and invalid data filtering;
in the step S4, an identity authentication algorithm center based on the individual behavior cohesiveness is built and comprises three algorithm module construction and two database construction; the three algorithm modules are respectively an individual behavior cohesiveness modeling algorithm module, a behavior cohesiveness probabilistic algorithm module and an authentication result output module based on threshold judgment; the two databases are respectively an individual behavior library and an attribute embedded library; the individual behavior cohesiveness modeling algorithm module comprises the steps of setting the dimension of an embedded vector space, defining a behavior cohesiveness vector, individual behavior cohesiveness and cohesiveness partial order, constructing an optimized objective function of the individual behavior cohesiveness, designing a parameter optimization method based on random gradient descent, performing model training, storing an individual behavior vector obtained by training into an individual behavior library, and storing an attribute embedded vector obtained by training into an attribute embedded library, wherein:
for behavior Attribute A j (J =1,2, \8230;, J) of any one of the attribute values a j It is mapped to a point in d-dimensional vector space, i.e. to a point in d-dimensional vector space
One behavior τ can be described as an embedded matrix m (τ) = [ a [ ] 1 ，a 2 ，...，a J ]Using the similarity between embedded matrix column vectors, a cohesive vector for behavior τ can be defined as h (τ) = (a) 1 ·a 2 ，a 1 ·a 3 ，...，a 1 ·a J ，a 2 ·a 3 ，...，a 2 ·a J ，...，a J-1 ·a J ) If the number of attributes describing a behavior is J, then its cohesiveness vector dimension is K = J. (J-1)/2, and accordingly, a K-dimension vector @isused>
Describing the behavior pattern of an individual i, namely a behavior vector of the individual, and defining the cohesiveness of the individual to the behavior on the basis of the behavior vector as follows:
the cohesiveness of an individual i to a behavior tau is defined as its behavior vector b i Inner product with cohesion vector h (τ), i.e.
For an individual i and a behavior τ p ，τ q E.g. T, cohesive partial order tau p ＞ i τ q Representing i pairs of behavior τ of individuals p Has a cohesion higher than its behavior τ q Cohesion of (d) if τ p ＞ i τ q Then by the action τ p Authenticating an individual i as being more legitimate than the behaviour τ q Then the individual's cohesive preference for behavior satisfies the condition:
if the individual's cohesive preference for behavior is treated as a random event, then the corresponding set of cohesive preference events is: omega i ＝{τ p ＞ i τ q |τ p ，τ q E.g. T, assuming that the cohesive preference events of the individual i are independent of each other, the joint probability of all preference events is expressed as: />
For preference event τ p ＞ i τ q The probability of its occurrence can be calculated using the following mathematical expression: />
Wherein the content of the first and second substances,
the function of the method is to convert any real number into a value in an interval (0-1), and further assume that cohesive preference events corresponding to different individuals are independent from each other, an optimized objective function is expressed as: />
Wherein I is the set of all individuals, and theta = { b = i ，v 1 ：，v 2 ：，...，v J : i ∈ I } represents the set of parameters of the model, then the optimal parameters can be obtained by maximizing the objective function: />
。
2. The identity authentication method based on the individual behavior cohesion, as claimed in claim 1, characterized in that, for optimizing the objective function, the objective function is converted into negative logarithmic form, i.e. the form of negative logarithm
Wherein it is present>
The parameter lambda controls the influence degree of the regular term, and after the objective function is converted, the optimized parameter can be obtained in the following mode: />
Model parameter optimization is carried out by adopting a random gradient descent method, and for any cohesiveness preference tau p ＞ i τ q In terms of this, the gradient of the objective function to the model parameters is represented as:
wherein, delta = - (1-sigma (b) i ·(h(τ p )-h(τ q ) )) is a constant that depends on the current iteration parameters of the model, a) p，s And a q，s Are respectively the behavior τ p And τ q The s (s =1, 2...., J) th column vector of the embedding matrix, according to the gradient formula, the iterative method of parameters in the training process is as follows:
where η is the learning rate used to control the span of the parameter in each step of the iteration process.
3. The identity authentication method based on the individual behavior cohesion as claimed in claim 2, wherein the primary key of the individual behavior library is an identification number representing an individual; the main key of the attribute embedded library is a uniform code containing a field name and field value information.
4. The identity authentication method based on the individual behavior cohesion as claimed in claim 3, wherein the behavior cohesion probabilistic algorithm module comprises obtaining behavior vectors of individuals by querying from an individual behavior library, obtaining corresponding vector representations by querying from an attribute embedding library according to field names and field values of the behaviors, calculating the cohesion vectors of the behaviors, obtaining individual behavior data by calculation according to the cohesion definition of the behaviors by the individuals, performing probabilistic processing on the individual behavior data, and outputting the legality probability of the identity authentication, wherein,
after the model training is finished, for any individual i, obtaining a behavior vector b representing the behavior pattern of the individual i i For arbitrary behavior τ =<a 1 ，a 2 ，...，a J >Obtaining an embedding matrix m (tau) of the individual behavior, further obtaining an aggregation vector h (tau) of the individual behavior, and correspondingly obtaining the cohesion of the individual behavior
As a basis for the legitimacy of an individual i authenticated with a behavior τ, however>
The value (b) is an arbitrary real number, has no fixed value range, and therefore cannot be directly used, and needs to be normalized, that is, the cohesion is based on the individual behavior>
This is translated into a probability of validity to authenticate an individual i with a behavior τ: />
The method comprises the following specific steps: />
A. For all behaviors except τ, [ tau ]' [ epsilon ] T, the behavior cohesiveness of the individual i is calculated separately
And their collection is denoted F i ；
B. Set counter C, traverse F i Behavior cohesion value of (1)
If->
Then the value of C is added with 1;
C. a probability of validity to authenticate an individual i with a behavior tau is calculated,
。
5. the identity authentication method based on the individual behavior cohesion as claimed in claim 4, wherein the authentication result output module based on the threshold judgment comprises a threshold setting module for comparing the validity probability of the output identity authentication with a threshold to determine whether the authentication is passed, wherein if the threshold is p, if so, the set threshold is p
The identity authentication is passed; on the contrary, if- >
The identity authentication is not passed.
6. The identity authentication system based on the individual behavior cohesiveness of the identity authentication method based on the individual behavior cohesiveness as claimed in any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110838416.0A CN113726736B (en) | 2021-07-23 | 2021-07-23 | Identity authentication method and system based on individual behavior cohesion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110838416.0A CN113726736B (en) | 2021-07-23 | 2021-07-23 | Identity authentication method and system based on individual behavior cohesion |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113726736A CN113726736A (en) | 2021-11-30 |
| CN113726736B true CN113726736B (en) | 2023-04-07 |
Family
ID=78673883
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110838416.0A Active CN113726736B (en) | 2021-07-23 | 2021-07-23 | Identity authentication method and system based on individual behavior cohesion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113726736B (en) |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111814128A (en) * | 2020-09-01 | 2020-10-23 | 北京远鉴信息技术有限公司 | Identity authentication method, device, equipment and storage medium based on fusion characteristics |
Family Cites Families (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105450412B (en) * | 2014-08-19 | 2019-04-30 | 阿里巴巴集团控股有限公司 | Identity identifying method and device |
| CN104202339B (en) * | 2014-09-24 | 2018-01-05 | 广西大学 | A kind of across cloud authentication service method based on user behavior |
| CN104318136A (en) * | 2014-09-29 | 2015-01-28 | 同济大学 | Modeling and analysis system for user keyboard key pressing activity mode and identity identification method thereof |
| CN105279405B (en) * | 2015-10-28 | 2018-06-26 | 同济大学 | Touchscreen user button behavior pattern is built and analysis system and its personal identification method |
| CN107194219A (en) * | 2017-06-16 | 2017-09-22 | 西安电子科技大学 | Intelligent terminal identity identifying method based on similarity |
| CN107819945B (en) * | 2017-10-30 | 2020-11-03 | 同济大学 | Handheld device browsing behavior authentication method and system integrating multiple factors |
| US11025419B2 (en) * | 2017-11-15 | 2021-06-01 | Alexander J. M. Van Der Velden | System for digital identity authentication and methods of use |
| CN108549806B (en) * | 2018-04-11 | 2020-01-31 | 西安电子科技大学 | Identity authentication method based on user sliding and clicking behaviors |
| CN112766292A (en) * | 2019-11-04 | 2021-05-07 | 中移(上海)信息通信科技有限公司 | Identity authentication method, device, equipment and storage medium |
| CN111917801A (en) * | 2020-08-18 | 2020-11-10 | 南京工业大学浦江学院 | Petri network-based user behavior authentication method in private cloud environment |
-
2021
- 2021-07-23 CN CN202110838416.0A patent/CN113726736B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111814128A (en) * | 2020-09-01 | 2020-10-23 | 北京远鉴信息技术有限公司 | Identity authentication method, device, equipment and storage medium based on fusion characteristics |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113726736A (en) | 2021-11-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106250858B (en) | Recognition method and system fusing multiple face recognition algorithms | |
| CN1892666A (en) | Method and apparatus for sequential authentication user | |
| WO2021135449A1 (en) | Deep reinforcement learning-based data classification method, apparatus, device, and medium | |
| CN112464117A (en) | Request processing method and device, computer equipment and storage medium | |
| CN109242250A (en) | A kind of user's behavior confidence level detection method based on Based on Entropy method and cloud model | |
| WO2021204086A1 (en) | Identity authentication method, and method and device for training identity authentication model | |
| CN109117668A (en) | A kind of identification authorization safety access method based on block chain building | |
| CN113553377B (en) | Data sharing method and device based on block chain and federal learning | |
| CN109753783A (en) | A kind of single-point logging method based on machine learning, device and computer readable storage medium | |
| CN113011895A (en) | Associated account sample screening method, device and equipment and computer storage medium | |
| CN113435505A (en) | Construction method and device for safe user portrait | |
| WO2021120587A1 (en) | Method and apparatus for retina classification based on oct, computer device, and storage medium | |
| JP2020107331A (en) | User verification method and apparatus using generalized user model | |
| CN113889262A (en) | Model-based data prediction method and device, computer equipment and storage medium | |
| US20210326475A1 (en) | Systems and method for evaluating identity disclosure risks in synthetic personal data | |
| CN114219596A (en) | Data processing method based on decision tree model and related equipment | |
| Sulayman et al. | User modeling via anomaly detection techniques for user authentication | |
| CN103970651A (en) | Software architecture safety assessment method based on module safety attributes | |
| CN113726736B (en) | Identity authentication method and system based on individual behavior cohesion | |
| CN117171786A (en) | Decentralizing federal learning method for resisting poisoning attack | |
| CN116777646A (en) | Artificial intelligence-based risk identification method, apparatus, device and storage medium | |
| CN111275059A (en) | Image processing method and device and computer readable storage medium | |
| Laskar et al. | Artificial Neural Networks and Gene Expression Programing based age estimation using facial features | |
| CN116405262A (en) | Network security access method, device, equipment and storage medium | |
| CN105678157B (en) | A kind of data property right protection system and method based on application environment identification |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |